Learn about evaluation sets
Evaluations tell you whether your Adobe Brand Concierge is answering correctly, staying in scope, and handling complex questions safely. This video walks through the three evaluation types — Functional, Out-of-Scope, and Safeguard — and shows how to build an evaluation set, run it, and review flagged results before you go live.
Who is this video for?
- Marketers and content owners preparing to launch
- Anyone validating concierge responses before go-live
- Teams responsible for concierge quality and safety
Video content
- The three evaluation types: Functional, Out-of-Scope, and Safeguard
- How to build an evaluation set — AI-generated questions or your own CSV upload
- How to run an evaluation and review the overall score and flagged questions
- How to compare the concierge’s actual response to the expected response for each flagged question
For documentation, see Brand Concierge help.
recommendation-more-help
brand-concierge-help-main-toc